LangChain Integration

You can use the WDK MCP Toolkit as a tool provider for LangChain agents in both Python and TypeScript. LangChain's MultiServerMCPClient spawns the MCP server as a subprocess and converts WDK tools into LangChain-compatible tools, giving your agent access to wallet operations, pricing, swaps, bridges, lending, and more.

This integration uses the serve CLI command, which starts a fully configured MCP server on stdio with no server script required.

The serve Command

The serve command provides zero-config MCP server startup so you don't need to write a server script:

npx @tetherto/wdk-mcp-toolkit serve

Pass WDK_SEED to enable wallet operations, or omit it to run with pricing tools only:

# With wallet operations
WDK_SEED="your twelve word seed phrase here" npx @tetherto/wdk-mcp-toolkit serve

# Pricing-only mode (no seed required)
npx @tetherto/wdk-mcp-toolkit serve

Default Chains

By default, serve enables three chains: Ethereum, Arbitrum, and Bitcoin. For each enabled chain it dynamically imports the required wallet package and skips any that aren't installed. You can change the enabled set with the WDK_CHAINS environment variable.

Built-in Registry

The command has built-in definitions for 13 chains and 4 protocol modules. When a chain is enabled and its package is installed, the wallet is registered automatically. Protocol modules are also auto-registered when their packages are installed and at least one of their target chains is enabled.

Quick Start

pip install langchain-mcp-adapters langgraph langchain-openai

import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

async def main():
    client = MultiServerMCPClient({
        "wdk": {
            "transport": "stdio",
            "command": "npx",
            "args": ["-y", "@tetherto/wdk-mcp-toolkit", "serve"],
            "env": {
                "WDK_SEED": "your twelve word seed phrase here",
                "WDK_MCP_ELICITATION": "false",
            },
        }
    })

    tools = await client.get_tools()
    agent = create_react_agent(ChatOpenAI(model="gpt-4o"), tools)

    result = await agent.ainvoke({
        "messages": [{"role": "user", "content": "What is my Ethereum address?"}]
    })
    print(result["messages"][-1].content)

    await client.close()

asyncio.run(main())

export OPENAI_API_KEY="sk-..."
python agent.py

Configuration

Environment Variables

Control serve behavior through environment variables:

Custom Config File

For chains or protocols not in the built-in defaults, create a wdk.config.json and pass its path via WDK_CONFIG:

WDK_CONFIG=./wdk.config.json WDK_SEED="..." npx @tetherto/wdk-mcp-toolkit serve

{
  "chains": {
    "zksync": {
      "module": "@myorg/wdk-wallet-zksync",
      "config": { "provider": "https://mainnet.era.zksync.io" }
    },
    "ethereum": {
      "config": { "provider": "https://my-private-rpc.com" }
    }
  },
  "protocols": [
    {
      "module": "@myorg/wdk-protocol-swap-custom",
      "label": "custom-swap",
      "type": "swap",
      "chains": ["zksync"]
    }
  ],
  "enabledChains": ["ethereum", "zksync", "bitcoin"]
}

LLM Provider Support

Both the Python and TypeScript examples support OpenAI and Anthropic. Set the corresponding environment variable and install the matching package:

Need Help?